Workplace Machine Learning Improves Accuracy, But Also Increases Human Workload, Study Shows

A new study from ESMT Berlin shows that the use of machine learning in the workplace always improves the accuracy of human decision-making, but in many cases it can also increase human cognitive effort during decision-making.

These findings come from research by Tamer Boyaci and Francis de Véricourt, professors of management science at ESMT Berlin, and Caner Canykmaz, a former postdoctoral fellow at ESMT and now assistant professor of operations management at Ojeguin University. It is a thing. Researchers wanted to explore how machine-based forecasting affects the decision-making processes and outcomes of human decision-makers. their paper is Business Administration.

Interestingly, the use of machines increases human workload the most when experts are cognitively constrained, such as experiencing time pressure or multitasking. However, situations where a decision maker experiences a high workload seem most appealing when he introduces AI just to alleviate some of this load. The study suggests that in this case, using AI to speed up the process may backfire, increasing human cognitive effort rather than decreasing it.

The researchers also found that while machine input always improves the overall accuracy of human decisions, it can also increase the likelihood of certain types of errors, such as false positives. In this study, we used machine learning models to compare decisions made by humans alone and those made by machines to identify differences in accuracy, propensity, and levels of cognitive effort by humans.

“Recently, the rapid adoption of AI technology by many organizations has raised concerns that AI will eventually replace humans in certain tasks,” said Professor de Bericourt. “But machines can greatly enhance human complementary strengths when used in tandem with human reason,” he says.

The researchers say their findings clearly demonstrate the value of human-machine collaboration for professionals. But even though machines can provide incredibly accurate information, humans often rely on humans to evaluate their own information before making decisions and compare machine prescriptions to their own conclusions. It should also be noted that it requires a lot of cognitive effort. Researchers say that when humans are under pressure to make decisions, the level of cognitive effort required increases.

“Machines, with their tremendous computing power, can perform certain tasks with incredible precision, whereas human decision-makers, by contrast, are flexible and adaptable, but have limited cognitive abilities. They are constrained and their skills complement each other,” says Professor Boyasi. “But humans need to be aware of the situations in which they use machines and understand when they are effective and when they are not.”

Using the example of doctors and patients, the researchers’ findings suggest that the use of machines improves overall diagnostic accuracy and reduces the number of misdiagnosed patients. However, when disease incidence is low and time is limited, the introduction of machines to assist physicians in diagnosing will increase patient misdiagnosis and require additional cognitive effort to resolve due to ambiguity. Therefore, diagnosis may require more human cognitive effort. The implementation machine may be the cause.

The researchers say their findings offer both hope and warning for those looking to introduce machines to their jobs. On the positive side, if the average accuracy improves and the machine tends to confirm that the input is rather predictable, all error rates decrease and it is “efficient” because it reduces human cognitive effort. will be

However, incorporating machine-based predictions into human decision-making is not always beneficial, either in terms of error reduction or amount of cognitive effort. In fact, introducing machines to improve the decision-making process is counterproductive because it can increase certain types of errors and increase the time and cognitive effort it takes to reach a decision. may become.

The findings highlight the significant impact machine-based predictions have on human judgments and decisions. These findings provide guidance on when and how machine input should be considered, and thus the design of human-machine collaboration.